Evaluation and Comparison of RFI Detection Algorithms

Anthropogenic Radio Frequency Interference (RFI) is an increasing problem in microwave remote sensing radiometry. Therefore, there is a growing interest on methods to detect and filter RFI. In this study, the performance of several different detection algorithms has been studied and compared to detect Continuous Wave (CW), QPSK modulated, and pulse modulated (with 0.1%, 1%, and 10% duty cycles) RFI. The mission scenario corresponds a spaceborne, polar-orbiting, conically scanning microwave radiometer. However, the qualitative results (e.g., the relative performances of algorithms) are applicable to other scenarios as well. It has been shown that RFI detection thresholds in 1 K range can be achieved in this scenario if complementary RFI detection algorithms can be incorporated in the system, e.g., in a digital RFI processor. Even 0.1 K level can be achieved for pulsed RFI with low duty cycles. Kurtosis and spectral kurtosis are effective in detecting pulsed RFI but not optimal in detecting constant envelope signals (such as CW or QPSK). Spectral Density Estimation and polarimetry, on the other hand, have a performance that is independent on the modulation or duty cycle of the RFI - they are sensitive to the average power.